diff options
author | Tom Gundersen <teg@jklm.no> | 2014-01-13 20:14:44 +0100 |
---|---|---|
committer | Tom Gundersen <teg@jklm.no> | 2014-01-13 21:06:13 +0100 |
commit | 0b54473e9b73d867ed7807507a0fb5adc8137b10 (patch) | |
tree | 0786f659f1b389adb5386636c4a1a9d49b7c2fab /src/libsystemd | |
parent | c813ca40c859ff8abc8bc6aabc3f1e896623eb67 (diff) |
libsystemd-rtnl: merge into libsystemd
Diffstat (limited to 'src/libsystemd')
-rw-r--r-- | src/libsystemd/rtnl-internal.h | 99 | ||||
-rw-r--r-- | src/libsystemd/rtnl-message.c | 887 | ||||
-rw-r--r-- | src/libsystemd/rtnl-util.c | 100 | ||||
-rw-r--r-- | src/libsystemd/rtnl-util.h | 36 | ||||
-rw-r--r-- | src/libsystemd/sd-rtnl.c | 870 | ||||
-rw-r--r-- | src/libsystemd/test-rtnl.c | 330 |
6 files changed, 2322 insertions, 0 deletions
diff --git a/src/libsystemd/rtnl-internal.h b/src/libsystemd/rtnl-internal.h new file mode 100644 index 0000000000..2e0b7b8f37 --- /dev/null +++ b/src/libsystemd/rtnl-internal.h @@ -0,0 +1,99 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright 2013 Tom Gundersen <teg@jklm.no> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <linux/netlink.h> + +#include "refcnt.h" +#include "prioq.h" +#include "list.h" + +#include "sd-rtnl.h" + +struct reply_callback { + sd_rtnl_message_handler_t callback; + void *userdata; + usec_t timeout; + uint64_t serial; + unsigned prioq_idx; +}; + +struct match_callback { + sd_rtnl_message_handler_t callback; + uint16_t type; + void *userdata; + + LIST_FIELDS(struct match_callback, match_callbacks); +}; + +struct sd_rtnl { + RefCount n_ref; + + int fd; + + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } sockaddr; + + sd_rtnl_message **rqueue; + unsigned rqueue_size; + + sd_rtnl_message **wqueue; + unsigned wqueue_size; + + bool processing:1; + + uint32_t serial; + + struct Prioq *reply_callbacks_prioq; + Hashmap *reply_callbacks; + + LIST_HEAD(struct match_callback, match_callbacks); + + pid_t original_pid; + + sd_event_source *io_event_source; + sd_event_source *time_event_source; + sd_event_source *exit_event_source; + sd_event *event; +}; + +#define RTNL_DEFAULT_TIMEOUT ((usec_t) (10 * USEC_PER_SEC)) + +#define RTNL_WQUEUE_MAX 1024 +#define RTNL_RQUEUE_MAX 64*1024 + +int message_new_synthetic_error(int error, uint32_t serial, sd_rtnl_message **ret); +uint32_t message_get_serial(sd_rtnl_message *m); +int message_seal(sd_rtnl *nl, sd_rtnl_message *m); + +bool message_type_is_link(uint16_t type); +bool message_type_is_addr(uint16_t type); +bool message_type_is_route(uint16_t type); + +int socket_write_message(sd_rtnl *nl, sd_rtnl_message *m); +int socket_read_message(sd_rtnl *nl, sd_rtnl_message **ret); + +/* Make sure callbacks don't destroy the rtnl connection */ +#define RTNL_DONT_DESTROY(rtnl) \ + _cleanup_sd_rtnl_unref_ _unused_ sd_rtnl *_dont_destroy_##rtnl = sd_rtnl_ref(rtnl) diff --git a/src/libsystemd/rtnl-message.c b/src/libsystemd/rtnl-message.c new file mode 100644 index 0000000000..517df61156 --- /dev/null +++ b/src/libsystemd/rtnl-message.c @@ -0,0 +1,887 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2013 Tom Gundersen <teg@jklm.no> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <linux/rtnetlink.h> +#include <netinet/in.h> +#include <netinet/ether.h> +#include <stdbool.h> +#include <unistd.h> + +#include "util.h" +#include "refcnt.h" + +#include "sd-rtnl.h" +#include "rtnl-internal.h" + +struct sd_rtnl_message { + RefCount n_ref; + + struct nlmsghdr *hdr; + size_t container_offset; /* offset from hdr to container start */ + size_t next_rta_offset; /* offset from hdr to next rta */ + + bool sealed:1; +}; + +#define CURRENT_CONTAINER(m) ((m)->container_offset ? (struct rtattr*)((uint8_t*)(m)->hdr + (m)->container_offset) : NULL) +#define NEXT_RTA(m) ((struct rtattr*)((uint8_t*)(m)->hdr + (m)->next_rta_offset)) +#define UPDATE_RTA(m, new) (m)->next_rta_offset = (uint8_t*)(new) - (uint8_t*)(m)->hdr; + +static int message_new(sd_rtnl_message **ret, size_t initial_size) { + sd_rtnl_message *m; + + assert_return(ret, -EINVAL); + assert_return(initial_size >= sizeof(struct nlmsghdr), -EINVAL); + + m = new0(sd_rtnl_message, 1); + if (!m) + return -ENOMEM; + + m->hdr = malloc0(initial_size); + if (!m->hdr) { + free(m); + return -ENOMEM; + } + + m->n_ref = REFCNT_INIT; + + m->hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK; + m->sealed = false; + + *ret = m; + + return 0; +} + +int message_new_synthetic_error(int error, uint32_t serial, sd_rtnl_message **ret) { + struct nlmsgerr *err; + int r; + + assert(error <= 0); + + r = message_new(ret, NLMSG_SPACE(sizeof(struct nlmsgerr))); + if (r < 0) + return r; + + (*ret)->hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr)); + (*ret)->hdr->nlmsg_type = NLMSG_ERROR; + (*ret)->hdr->nlmsg_seq = serial; + + err = NLMSG_DATA((*ret)->hdr); + + err->error = error; + + return 0; +} + +bool message_type_is_route(uint16_t type) { + switch (type) { + case RTM_NEWROUTE: + case RTM_GETROUTE: + case RTM_DELROUTE: + return true; + default: + return false; + } +} + +bool message_type_is_link(uint16_t type) { + switch (type) { + case RTM_NEWLINK: + case RTM_SETLINK: + case RTM_GETLINK: + case RTM_DELLINK: + return true; + default: + return false; + } +} + +bool message_type_is_addr(uint16_t type) { + switch (type) { + case RTM_NEWADDR: + case RTM_GETADDR: + case RTM_DELADDR: + return true; + default: + return false; + } +} + +int sd_rtnl_message_route_set_dst_prefixlen(sd_rtnl_message *m, unsigned char prefixlen) { + struct rtmsg *rtm; + + rtm = NLMSG_DATA(m->hdr); + + rtm->rtm_dst_len = prefixlen; + + return 0; +} + +int sd_rtnl_message_route_new(uint16_t nlmsg_type, unsigned char rtm_family, + sd_rtnl_message **ret) { + struct rtmsg *rtm; + int r; + + assert_return(message_type_is_route(nlmsg_type), -EINVAL); + assert_return(rtm_family == AF_INET || rtm_family == AF_INET6, -EINVAL); + assert_return(ret, -EINVAL); + + r = message_new(ret, NLMSG_SPACE(sizeof(struct rtmsg))); + if (r < 0) + return r; + + (*ret)->hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg)); + (*ret)->hdr->nlmsg_type = nlmsg_type; + if (nlmsg_type == RTM_NEWROUTE) + (*ret)->hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL; + + rtm = NLMSG_DATA((*ret)->hdr); + + UPDATE_RTA(*ret, RTM_RTA(rtm)); + + rtm->rtm_family = rtm_family; + rtm->rtm_scope = RT_SCOPE_UNIVERSE; + rtm->rtm_type = RTN_UNICAST; + rtm->rtm_table = RT_TABLE_MAIN; + rtm->rtm_protocol = RTPROT_BOOT; + + return 0; +} + +int sd_rtnl_message_link_set_flags(sd_rtnl_message *m, unsigned flags) { + struct ifinfomsg *ifi; + + ifi = NLMSG_DATA(m->hdr); + + ifi->ifi_flags = flags; + + return 0; +} + +int sd_rtnl_message_link_set_type(sd_rtnl_message *m, unsigned type) { + struct ifinfomsg *ifi; + + ifi = NLMSG_DATA(m->hdr); + + ifi->ifi_type = type; + + return 0; +} + +int sd_rtnl_message_link_new(uint16_t nlmsg_type, int index, sd_rtnl_message **ret) { + struct ifinfomsg *ifi; + int r; + + assert_return(message_type_is_link(nlmsg_type), -EINVAL); + assert_return(nlmsg_type == RTM_NEWLINK || index > 0, -EINVAL); + assert_return(ret, -EINVAL); + + r = message_new(ret, NLMSG_SPACE(sizeof(struct ifinfomsg))); + if (r < 0) + return r; + + (*ret)->hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)); + (*ret)->hdr->nlmsg_type = nlmsg_type; + if (nlmsg_type == RTM_NEWLINK) + (*ret)->hdr->nlmsg_flags |= NLM_F_CREATE; + + ifi = NLMSG_DATA((*ret)->hdr); + + ifi->ifi_family = AF_UNSPEC; + ifi->ifi_index = index; + ifi->ifi_change = 0xffffffff; + + UPDATE_RTA(*ret, IFLA_RTA(ifi)); + + return 0; +} + +int sd_rtnl_message_addr_new(uint16_t nlmsg_type, int index, unsigned char family, unsigned char prefixlen, unsigned char flags, unsigned char scope, sd_rtnl_message **ret) { + struct ifaddrmsg *ifa; + int r; + + assert_return(message_type_is_addr(nlmsg_type), -EINVAL); + assert_return(index > 0, -EINVAL); + assert_return(ret, -EINVAL); + + r = message_new(ret, NLMSG_SPACE(sizeof(struct ifaddrmsg))); + if (r < 0) + return r; + + (*ret)->hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg)); + (*ret)->hdr->nlmsg_type = nlmsg_type; + + ifa = NLMSG_DATA((*ret)->hdr); + + ifa->ifa_family = family; + ifa->ifa_prefixlen = prefixlen; + ifa->ifa_flags = flags; + ifa->ifa_scope = scope; + ifa->ifa_index = index; + + UPDATE_RTA(*ret, IFA_RTA(ifa)); + + return 0; +} + +sd_rtnl_message *sd_rtnl_message_ref(sd_rtnl_message *m) { + if (m) + assert_se(REFCNT_INC(m->n_ref) >= 2); + + return m; +} + +sd_rtnl_message *sd_rtnl_message_unref(sd_rtnl_message *m) { + if (m && REFCNT_DEC(m->n_ref) <= 0) { + free(m->hdr); + free(m); + } + + return NULL; +} + +int sd_rtnl_message_get_type(sd_rtnl_message *m, uint16_t *type) { + assert_return(m, -EINVAL); + assert_return(type, -EINVAL); + + *type = m->hdr->nlmsg_type; + + return 0; +} + +int sd_rtnl_message_link_get_ifindex(sd_rtnl_message *m, int *ifindex) { + struct ifinfomsg *ifi; + + assert_return(m, -EINVAL); + assert_return(m->hdr, -EINVAL); + assert_return(message_type_is_link(m->hdr->nlmsg_type), -EINVAL); + assert_return(ifindex, -EINVAL); + + ifi = NLMSG_DATA(m->hdr); + + *ifindex = ifi->ifi_index; + + return 0; +} + +int sd_rtnl_message_link_get_flags(sd_rtnl_message *m, unsigned *flags) { + struct ifinfomsg *ifi; + + assert_return(m, -EINVAL); + assert_return(m->hdr, -EINVAL); + assert_return(message_type_is_link(m->hdr->nlmsg_type), -EINVAL); + assert_return(flags, -EINVAL); + + ifi = NLMSG_DATA(m->hdr); + + *flags = ifi->ifi_flags; + + return 0; +} + +/* If successful the updated message will be correctly aligned, if + unsuccessful the old message is untouched. */ +static int add_rtattr(sd_rtnl_message *m, unsigned short type, const void *data, size_t data_length) { + uint32_t rta_length, message_length; + struct nlmsghdr *new_hdr; + struct rtattr *rta; + char *padding; + + assert(m); + assert(m->hdr); + assert(NLMSG_ALIGN(m->hdr->nlmsg_len) == m->hdr->nlmsg_len); + assert(!data || data_length > 0); + + /* get the size of the new rta attribute (with padding at the end) */ + rta_length = RTA_LENGTH(data_length); + + /* get the new message size (with padding at the end) */ + message_length = m->hdr->nlmsg_len + RTA_ALIGN(rta_length); + + /* realloc to fit the new attribute */ + new_hdr = realloc(m->hdr, message_length); + if (!new_hdr) + return -ENOMEM; + m->hdr = new_hdr; + + /* get pointer to the attribute we are about to add */ + rta = (struct rtattr *) ((uint8_t *) m->hdr + m->hdr->nlmsg_len); + + /* if we are inside a container, extend it */ + if (CURRENT_CONTAINER(m)) + CURRENT_CONTAINER(m)->rta_len += message_length - m->hdr->nlmsg_len; + + /* fill in the attribute */ + rta->rta_type = type; + rta->rta_len = rta_length; + if (!data) { + /* this is the start of a new container */ + m->container_offset = m->hdr->nlmsg_len; + } else { + /* we don't deal with the case where the user lies about the type + * and gives us too little data (so don't do that) + */ + padding = mempcpy(RTA_DATA(rta), data, data_length); + /* make sure also the padding at the end of the message is initialized */ + memzero(padding, + (uint8_t *) m->hdr + message_length - (uint8_t *) padding); + } + + /* update message size */ + m->hdr->nlmsg_len = message_length; + + return 0; +} + +int sd_rtnl_message_append_string(sd_rtnl_message *m, unsigned short type, const char *data) { + uint16_t rtm_type; + int r; + + assert_return(m, -EINVAL); + assert_return(data, -EINVAL); + + r = sd_rtnl_message_get_type(m, &rtm_type); + if (r < 0) + return r; + + /* check that the type is correct */ + switch (rtm_type) { + case RTM_NEWLINK: + case RTM_SETLINK: + case RTM_GETLINK: + case RTM_DELLINK: + if (CURRENT_CONTAINER(m)) { + if (CURRENT_CONTAINER(m)->rta_type != IFLA_LINKINFO || + type != IFLA_INFO_KIND) + return -ENOTSUP; + } else { + switch (type) { + case IFLA_IFNAME: + case IFLA_IFALIAS: + case IFLA_QDISC: + break; + default: + return -ENOTSUP; + } + } + break; + case RTM_NEWADDR: + case RTM_GETADDR: + case RTM_DELADDR: + if (type != IFA_LABEL) + return -ENOTSUP; + break; + default: + return -ENOTSUP; + } + + r = add_rtattr(m, type, data, strlen(data) + 1); + if (r < 0) + return r; + + return 0; +} + +int sd_rtnl_message_append_u32(sd_rtnl_message *m, unsigned short type, uint32_t data) { + uint16_t rtm_type; + int r; + + assert_return(m, -EINVAL); + + r = sd_rtnl_message_get_type(m, &rtm_type); + if (r < 0) + return r; + + /* check that the type is correct */ + switch (rtm_type) { + case RTM_NEWLINK: + case RTM_SETLINK: + case RTM_GETLINK: + case RTM_DELLINK: + switch (type) { + case IFLA_MASTER: + case IFLA_MTU: + case IFLA_LINK: + break; + default: + return -ENOTSUP; + } + break; + case RTM_NEWROUTE: + case RTM_GETROUTE: + case RTM_DELROUTE: + switch (type) { + case RTA_TABLE: + case RTA_PRIORITY: + case RTA_IIF: + case RTA_OIF: + break; + default: + return -ENOTSUP; + } + break; + default: + return -ENOTSUP; + } + + r = add_rtattr(m, type, &data, sizeof(uint32_t)); + if (r < 0) + return r; + + return 0; +} + +int sd_rtnl_message_append_in_addr(sd_rtnl_message *m, unsigned short type, const struct in_addr *data) { + struct ifaddrmsg *ifa; + struct rtmsg *rtm; + uint16_t rtm_type; + int r; + + assert_return(m, -EINVAL); + assert_return(data, -EINVAL); + + r = sd_rtnl_message_get_type(m, &rtm_type); + if (r < 0) + return r; + + /* check that the type is correct */ + switch (rtm_type) { + case RTM_NEWADDR: + case RTM_GETADDR: + case RTM_DELADDR: + switch (type) { + case IFA_ADDRESS: + case IFA_LOCAL: + case IFA_BROADCAST: + case IFA_ANYCAST: + ifa = NLMSG_DATA(m->hdr); + + if (ifa->ifa_family != AF_INET) + return -EINVAL; + + break; + default: + return -ENOTSUP; + } + break; + case RTM_NEWROUTE: + case RTM_GETROUTE: + case RTM_DELROUTE: + switch (type) { + case RTA_DST: + case RTA_SRC: + case RTA_GATEWAY: + rtm = NLMSG_DATA(m->hdr); + + if (rtm->rtm_family != AF_INET) + return -EINVAL; + + break; + default: + return -ENOTSUP; + } + break; + default: + return -ENOTSUP; + } + + r = add_rtattr(m, type, data, sizeof(struct in_addr)); + if (r < 0) + return r; + + return 0; +} + +int sd_rtnl_message_append_in6_addr(sd_rtnl_message *m, unsigned short type, const struct in6_addr *data) { + struct ifaddrmsg *ifa; + struct rtmsg *rtm; + uint16_t rtm_type; + int r; + + assert_return(m, -EINVAL); + assert_return(data, -EINVAL); + + r = sd_rtnl_message_get_type(m, &rtm_type); + if (r < 0) + return r; + + /* check that the type is correct */ + switch (rtm_type) { + case RTM_NEWADDR: + case RTM_GETADDR: + case RTM_DELADDR: + switch (type) { + case IFA_ADDRESS: + case IFA_LOCAL: + case IFA_BROADCAST: + case IFA_ANYCAST: + ifa = NLMSG_DATA(m->hdr); + + if (ifa->ifa_family != AF_INET6) + return -EINVAL; + + break; + default: + return -ENOTSUP; + } + break; + case RTM_NEWROUTE: + case RTM_GETROUTE: + case RTM_DELROUTE: + switch (type) { + case RTA_DST: + case RTA_SRC: + case RTA_GATEWAY: + rtm = NLMSG_DATA(m->hdr); + + if (rtm->rtm_family != AF_INET6) + return -EINVAL; + + break; + default: + return -ENOTSUP; + } + default: + return -ENOTSUP; + } + + r = add_rtattr(m, type, data, sizeof(struct in6_addr)); + if (r < 0) + return r; + + return 0; +} + +int sd_rtnl_message_append_ether_addr(sd_rtnl_message *m, unsigned short type, const struct ether_addr *data) { + uint16_t rtm_type; + int r; + + assert_return(m, -EINVAL); + assert_return(data, -EINVAL); + + sd_rtnl_message_get_type(m, &rtm_type); + + switch (rtm_type) { + case RTM_NEWLINK: + case RTM_SETLINK: + case RTM_DELLINK: + case RTM_GETLINK: + switch (type) { + case IFLA_ADDRESS: + case IFLA_BROADCAST: + break; + default: + return -ENOTSUP; + } + break; + default: + return -ENOTSUP; + } + + r = add_rtattr(m, type, data, ETH_ALEN); + if (r < 0) + return r; + + return 0; +} + +int sd_rtnl_message_open_container(sd_rtnl_message *m, unsigned short type) { + uint16_t rtm_type; + + assert_return(m, -EINVAL); + assert_return(!CURRENT_CONTAINER(m), -EINVAL); + + sd_rtnl_message_get_type(m, &rtm_type); + + if (message_type_is_link(rtm_type)) { + if (type == IFLA_LINKINFO) + return add_rtattr(m, type, NULL, 0); + else + return -ENOTSUP; + } else + return -ENOTSUP; + + return 0; +} + +int sd_rtnl_message_close_container(sd_rtnl_message *m) { + assert_return(m, -EINVAL); + assert_return(CURRENT_CONTAINER(m), -EINVAL); + + m->container_offset = 0; + + return 0; +} + +int sd_rtnl_message_read(sd_rtnl_message *m, unsigned short *type, void **data) { + size_t remaining_size; + uint16_t rtm_type; + int r; + + assert(m); + assert(m->next_rta_offset); + assert(type); + assert(data); + + remaining_size = m->hdr->nlmsg_len - m->next_rta_offset; + + if (!RTA_OK(NEXT_RTA(m), remaining_size)) + return 0; + + /* make sure we don't try to read a container + * TODO: add support for entering containers for reading */ + r = sd_rtnl_message_get_type(m, &rtm_type); + if (r < 0) + return r; + + if (message_type_is_link(rtm_type) && + NEXT_RTA(m)->rta_type == IFLA_LINKINFO) + return -EINVAL; + + *data = RTA_DATA(NEXT_RTA(m)); + *type = NEXT_RTA(m)->rta_type; + + UPDATE_RTA(m, RTA_NEXT(NEXT_RTA(m), remaining_size)); + + return 1; +} + +uint32_t message_get_serial(sd_rtnl_message *m) { + assert(m); + assert(m->hdr); + + return m->hdr->nlmsg_seq; +} + +int sd_rtnl_message_get_errno(sd_rtnl_message *m) { + struct nlmsgerr *err; + + assert_return(m, -EINVAL); + assert_return(m->hdr, -EINVAL); + + if (m->hdr->nlmsg_type != NLMSG_ERROR) + return 0; + + err = NLMSG_DATA(m->hdr); + + return err->error; +} + +int message_seal(sd_rtnl *nl, sd_rtnl_message *m) { + assert(nl); + assert(m); + assert(m->hdr); + + if (m->sealed) + return -EPERM; + + m->hdr->nlmsg_seq = nl->serial++; + m->sealed = true; + + return 0; +} + +static int message_receive_need(sd_rtnl *rtnl, size_t *need) { + assert(rtnl); + assert(need); + + /* ioctl(rtnl->fd, FIONREAD, &need) + Does not appear to work on netlink sockets. libnl uses + MSG_PEEK instead. I don't know if that is worth the + extra roundtrip. + + For now we simply use the maximum message size the kernel + may use (NLMSG_GOODSIZE), and then realloc to the actual + size after reading the message (hence avoiding huge memory + usage in case many small messages are kept around) */ + *need = page_size(); + if (*need > 8192UL) + *need = 8192UL; + + return 0; +} + +/* returns the number of bytes sent, or a negative error code */ +int socket_write_message(sd_rtnl *nl, sd_rtnl_message *m) { + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } addr = { + .nl.nl_family = AF_NETLINK, + }; + ssize_t k; + + assert(nl); + assert(m); + assert(m->hdr); + + k = sendto(nl->fd, m->hdr, m->hdr->nlmsg_len, + 0, &addr.sa, sizeof(addr)); + if (k < 0) + return (errno == EAGAIN) ? 0 : -errno; + + return k; +} + +/* On success, the number of bytes received is returned and *ret points to the received message + * which has a valid header and the correct size. + * If nothing useful was received 0 is returned. + * On failure, a negative error code is returned. + */ +int socket_read_message(sd_rtnl *nl, sd_rtnl_message **ret) { + sd_rtnl_message *m; + union { + struct sockaddr sa; + struct sockaddr_nl nl; + } addr; + socklen_t addr_len; + int r; + ssize_t k; + size_t need; + + assert(nl); + assert(ret); + + r = message_receive_need(nl, &need); + if (r < 0) + return r; + + r = message_new(&m, need); + if (r < 0) + return r; + + addr_len = sizeof(addr); + + k = recvfrom(nl->fd, m->hdr, need, + 0, &addr.sa, &addr_len); + if (k < 0) + k = (errno == EAGAIN) ? 0 : -errno; /* no data */ + else if (k == 0) + k = -ECONNRESET; /* connection was closed by the kernel */ + else if (addr_len != sizeof(addr.nl) || + addr.nl.nl_family != AF_NETLINK) + k = -EIO; /* not a netlink message */ + else if (addr.nl.nl_pid != 0) + k = 0; /* not from the kernel */ + else if ((size_t) k < sizeof(struct nlmsghdr) || + (size_t) k < m->hdr->nlmsg_len) + k = -EIO; /* too small (we do accept too big though) */ + else if (m->hdr->nlmsg_pid && m->hdr->nlmsg_pid != nl->sockaddr.nl.nl_pid) + k = 0; /* not broadcast and not for us */ + + if (k > 0) + switch (m->hdr->nlmsg_type) { + struct ifinfomsg *ifi; + struct ifaddrmsg *ifa; + struct rtmsg *rtm; + + /* check that the size matches the message type */ + case NLMSG_ERROR: + if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) + k = -EIO; + break; + case RTM_NEWLINK: + case RTM_SETLINK: + case RTM_DELLINK: + case RTM_GETLINK: + if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct ifinfomsg))) + k = -EIO; + else { + ifi = NLMSG_DATA(m->hdr); + UPDATE_RTA(m, IFLA_RTA(ifi)); + } + break; + case RTM_NEWADDR: + case RTM_DELADDR: + case RTM_GETADDR: + if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct ifaddrmsg))) + k = -EIO; + else { + ifa = NLMSG_DATA(m->hdr); + UPDATE_RTA(m, IFA_RTA(ifa)); + } + break; + case RTM_NEWROUTE: + case RTM_DELROUTE: + case RTM_GETROUTE: + if (m->hdr->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtmsg))) + k = -EIO; + else { + rtm = NLMSG_DATA(m->hdr); + UPDATE_RTA(m, RTM_RTA(rtm)); + } + break; + case NLMSG_NOOP: + k = 0; + break; + default: + k = 0; /* ignoring message of unknown type */ + } + + if (k <= 0) + sd_rtnl_message_unref(m); + else { + /* we probably allocated way too much memory, give it back */ + m->hdr = realloc(m->hdr, m->hdr->nlmsg_len); + *ret = m; + } + + return k; +} + +int sd_rtnl_message_rewind(sd_rtnl_message *m) { + struct ifinfomsg *ifi; + struct ifaddrmsg *ifa; + struct rtmsg *rtm; + + assert_return(m, -EINVAL); + assert_return(m->hdr, -EINVAL); + + switch(m->hdr->nlmsg_type) { + case RTM_NEWLINK: + case RTM_SETLINK: + case RTM_GETLINK: + case RTM_DELLINK: + ifi = NLMSG_DATA(m->hdr); + UPDATE_RTA(m, IFLA_RTA(ifi)); + + break; + case RTM_NEWADDR: + case RTM_GETADDR: + case RTM_DELADDR: + ifa = NLMSG_DATA(m->hdr); + UPDATE_RTA(m, IFA_RTA(ifa)); + + break; + case RTM_NEWROUTE: + case RTM_GETROUTE: + case RTM_DELROUTE: + rtm = NLMSG_DATA(m->hdr); + UPDATE_RTA(m, RTM_RTA(rtm)); + + break; + default: + return -ENOTSUP; + } + + return 0; +} diff --git a/src/libsystemd/rtnl-util.c b/src/libsystemd/rtnl-util.c new file mode 100644 index 0000000000..dfc0050def --- /dev/null +++ b/src/libsystemd/rtnl-util.c @@ -0,0 +1,100 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright (C) 2013 Tom Gundersen <teg@jklm.no> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <linux/rtnetlink.h> +#include <netinet/ether.h> + +#include "sd-rtnl.h" + +#include "rtnl-util.h" + +int rtnl_set_link_name(sd_rtnl *rtnl, int ifindex, const char *name) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *message = NULL; + int r; + + assert(rtnl); + assert(ifindex > 0); + assert(name); + + r = sd_rtnl_message_link_new(RTM_SETLINK, ifindex, &message); + if (r < 0) + return r; + + r = sd_rtnl_message_append_string(message, IFLA_IFNAME, name); + if (r < 0) + return r; + + r = sd_rtnl_call(rtnl, message, 0, NULL); + if (r < 0) + return r; + + return 0; +} + +int rtnl_set_link_properties(sd_rtnl *rtnl, int ifindex, const char *alias, + const struct ether_addr *mac, unsigned mtu) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *message = NULL; + bool need_update = false; + int r; + + assert(rtnl); + assert(ifindex > 0); + + if (!alias && !mac && mtu == 0) + return 0; + + r = sd_rtnl_message_link_new(RTM_SETLINK, ifindex, &message); + if (r < 0) + return r; + + if (alias) { + r = sd_rtnl_message_append_string(message, IFLA_IFALIAS, alias); + if (r < 0) + return r; + + need_update = true; + + } + + if (mac) { + r = sd_rtnl_message_append_ether_addr(message, IFLA_ADDRESS, mac); + if (r < 0) + return r; + + need_update = true; + } + + if (mtu > 0) { + r = sd_rtnl_message_append_u32(message, IFLA_MTU, mtu); + if (r < 0) + return r; + + need_update = true; + } + + if (need_update) { + r = sd_rtnl_call(rtnl, message, 0, NULL); + if (r < 0) + return r; + } + + return 0; +} diff --git a/src/libsystemd/rtnl-util.h b/src/libsystemd/rtnl-util.h new file mode 100644 index 0000000000..013002dd60 --- /dev/null +++ b/src/libsystemd/rtnl-util.h @@ -0,0 +1,36 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +#pragma once + +/*** + This file is part of systemd. + + Copyright (C) 2013 Tom Gundersen <teg@jklm.no> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <netinet/ether.h> + +#include "util.h" +#include "sd-rtnl.h" + +int rtnl_set_link_name(sd_rtnl *rtnl, int ifindex, const char *name); +int rtnl_set_link_properties(sd_rtnl *rtnl, int ifindex, const char *alias, const struct ether_addr *mac, unsigned mtu); + +DEFINE_TRIVIAL_CLEANUP_FUNC(sd_rtnl*, sd_rtnl_unref); +DEFINE_TRIVIAL_CLEANUP_FUNC(sd_rtnl_message*, sd_rtnl_message_unref); + +#define _cleanup_sd_rtnl_unref_ _cleanup_(sd_rtnl_unrefp) +#define _cleanup_sd_rtnl_message_unref_ _cleanup_(sd_rtnl_message_unrefp) diff --git a/src/libsystemd/sd-rtnl.c b/src/libsystemd/sd-rtnl.c new file mode 100644 index 0000000000..08b82ab2a9 --- /dev/null +++ b/src/libsystemd/sd-rtnl.c @@ -0,0 +1,870 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2013 Tom Gundersen <teg@jklm.no> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <sys/socket.h> +#include <poll.h> + +#include "macro.h" +#include "util.h" +#include "hashmap.h" + +#include "sd-rtnl.h" +#include "rtnl-internal.h" +#include "rtnl-util.h" + +static int sd_rtnl_new(sd_rtnl **ret) { + sd_rtnl *rtnl; + + assert_return(ret, -EINVAL); + + rtnl = new0(sd_rtnl, 1); + if (!rtnl) + return -ENOMEM; + + rtnl->n_ref = REFCNT_INIT; + + rtnl->fd = -1; + + rtnl->sockaddr.nl.nl_family = AF_NETLINK; + + rtnl->original_pid = getpid(); + + LIST_HEAD_INIT(rtnl->match_callbacks); + + /* We guarantee that wqueue always has space for at least + * one entry */ + rtnl->wqueue = new(sd_rtnl_message*, 1); + if (!rtnl->wqueue) { + free(rtnl); + return -ENOMEM; + } + + *ret = rtnl; + return 0; +} + +static bool rtnl_pid_changed(sd_rtnl *rtnl) { + assert(rtnl); + + /* We don't support people creating an rtnl connection and + * keeping it around over a fork(). Let's complain. */ + + return rtnl->original_pid != getpid(); +} + +int sd_rtnl_open(uint32_t groups, sd_rtnl **ret) { + _cleanup_sd_rtnl_unref_ sd_rtnl *rtnl = NULL; + socklen_t addrlen; + int r; + + assert_return(ret, -EINVAL); + + r = sd_rtnl_new(&rtnl); + if (r < 0) + return r; + + rtnl->fd = socket(PF_NETLINK, SOCK_RAW|SOCK_CLOEXEC|SOCK_NONBLOCK, NETLINK_ROUTE); + if (rtnl->fd < 0) + return -errno; + + rtnl->sockaddr.nl.nl_groups = groups; + + addrlen = sizeof(rtnl->sockaddr); + + r = bind(rtnl->fd, &rtnl->sockaddr.sa, addrlen); + if (r < 0) + return -errno; + + r = getsockname(rtnl->fd, &rtnl->sockaddr.sa, &addrlen); + if (r < 0) + return r; + + *ret = rtnl; + rtnl = NULL; + + return 0; +} + +sd_rtnl *sd_rtnl_ref(sd_rtnl *rtnl) { + if (rtnl) + assert_se(REFCNT_INC(rtnl->n_ref) >= 2); + + return rtnl; +} + +sd_rtnl *sd_rtnl_unref(sd_rtnl *rtnl) { + + if (rtnl && REFCNT_DEC(rtnl->n_ref) <= 0) { + struct match_callback *f; + unsigned i; + + for (i = 0; i < rtnl->rqueue_size; i++) + sd_rtnl_message_unref(rtnl->rqueue[i]); + free(rtnl->rqueue); + + for (i = 0; i < rtnl->wqueue_size; i++) + sd_rtnl_message_unref(rtnl->wqueue[i]); + free(rtnl->wqueue); + + hashmap_free_free(rtnl->reply_callbacks); + prioq_free(rtnl->reply_callbacks_prioq); + + while ((f = rtnl->match_callbacks)) { + LIST_REMOVE(match_callbacks, rtnl->match_callbacks, f); + free(f); + } + + if (rtnl->fd >= 0) + close_nointr_nofail(rtnl->fd); + + free(rtnl); + } + + return NULL; +} + +int sd_rtnl_send(sd_rtnl *nl, + sd_rtnl_message *message, + uint32_t *serial) { + int r; + + assert_return(nl, -EINVAL); + assert_return(!rtnl_pid_changed(nl), -ECHILD); + assert_return(message, -EINVAL); + + r = message_seal(nl, message); + if (r < 0) + return r; + + if (nl->wqueue_size <= 0) { + /* send directly */ + r = socket_write_message(nl, message); + if (r < 0) + return r; + else if (r == 0) { + /* nothing was sent, so let's put it on + * the queue */ + nl->wqueue[0] = sd_rtnl_message_ref(message); + nl->wqueue_size = 1; + } + } else { + sd_rtnl_message **q; + + /* append to queue */ + if (nl->wqueue_size >= RTNL_WQUEUE_MAX) + return -ENOBUFS; + + q = realloc(nl->wqueue, sizeof(sd_rtnl_message*) * (nl->wqueue_size + 1)); + if (!q) + return -ENOMEM; + + nl->wqueue = q; + q[nl->wqueue_size ++] = sd_rtnl_message_ref(message); + } + + if (serial) + *serial = message_get_serial(message); + + return 1; +} + +static int dispatch_rqueue(sd_rtnl *rtnl, sd_rtnl_message **message) { + sd_rtnl_message *z = NULL; + int r; + + assert(rtnl); + assert(message); + + if (rtnl->rqueue_size > 0) { + /* Dispatch a queued message */ + + *message = rtnl->rqueue[0]; + rtnl->rqueue_size --; + memmove(rtnl->rqueue, rtnl->rqueue + 1, sizeof(sd_rtnl_message*) * rtnl->rqueue_size); + + return 1; + } + + /* Try to read a new message */ + r = socket_read_message(rtnl, &z); + if (r < 0) + return r; + if (r == 0) + return 0; + + *message = z; + + return 1; +} + +static int dispatch_wqueue(sd_rtnl *rtnl) { + int r, ret = 0; + + assert(rtnl); + + while (rtnl->wqueue_size > 0) { + r = socket_write_message(rtnl, rtnl->wqueue[0]); + if (r < 0) + return r; + else if (r == 0) + /* Didn't do anything this time */ + return ret; + else { + /* see equivalent in sd-bus.c */ + sd_rtnl_message_unref(rtnl->wqueue[0]); + rtnl->wqueue_size --; + memmove(rtnl->wqueue, rtnl->wqueue + 1, sizeof(sd_rtnl_message*) * rtnl->wqueue_size); + + ret = 1; + } + } + + return ret; +} + +static int process_timeout(sd_rtnl *rtnl) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *m = NULL; + struct reply_callback *c; + usec_t n; + int r; + + assert(rtnl); + + c = prioq_peek(rtnl->reply_callbacks_prioq); + if (!c) + return 0; + + n = now(CLOCK_MONOTONIC); + if (c->timeout > n) + return 0; + + r = message_new_synthetic_error(-ETIMEDOUT, c->serial, &m); + if (r < 0) + return r; + + assert_se(prioq_pop(rtnl->reply_callbacks_prioq) == c); + hashmap_remove(rtnl->reply_callbacks, &c->serial); + + r = c->callback(rtnl, m, c->userdata); + free(c); + + return r < 0 ? r : 1; +} + +static int process_reply(sd_rtnl *rtnl, sd_rtnl_message *m) { + struct reply_callback *c; + uint64_t serial; + int r; + + assert(rtnl); + assert(m); + + serial = message_get_serial(m); + c = hashmap_remove(rtnl->reply_callbacks, &serial); + if (!c) + return 0; + + if (c->timeout != 0) + prioq_remove(rtnl->reply_callbacks_prioq, c, &c->prioq_idx); + + r = c->callback(rtnl, m, c->userdata); + free(c); + + return r; +} + +static int process_match(sd_rtnl *rtnl, sd_rtnl_message *m) { + struct match_callback *c; + uint16_t type; + int r; + + assert(rtnl); + assert(m); + + r = sd_rtnl_message_get_type(m, &type); + if (r < 0) + return r; + + LIST_FOREACH(match_callbacks, c, rtnl->match_callbacks) { + if (type == c->type) { + r = c->callback(rtnl, m, c->userdata); + if (r != 0) + return r; + } + } + + return 0; +} + +static int process_running(sd_rtnl *rtnl, sd_rtnl_message **ret) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *m = NULL; + int r; + + assert(rtnl); + + r = process_timeout(rtnl); + if (r != 0) + goto null_message; + + r = dispatch_wqueue(rtnl); + if (r != 0) + goto null_message; + + r = dispatch_rqueue(rtnl, &m); + if (r < 0) + return r; + if (!m) + goto null_message; + + r = process_reply(rtnl, m); + if (r != 0) + goto null_message; + + r = process_match(rtnl, m); + if (r != 0) + goto null_message; + + if (ret) { + *ret = m; + m = NULL; + + return 1; + } + + return 1; + +null_message: + if (r >= 0 && ret) + *ret = NULL; + + return r; +} + +int sd_rtnl_process(sd_rtnl *rtnl, sd_rtnl_message **ret) { + RTNL_DONT_DESTROY(rtnl); + int r; + + assert_return(rtnl, -EINVAL); + assert_return(!rtnl_pid_changed(rtnl), -ECHILD); + assert_return(!rtnl->processing, -EBUSY); + + rtnl->processing = true; + r = process_running(rtnl, ret); + rtnl->processing = false; + + return r; +} + +static usec_t calc_elapse(uint64_t usec) { + if (usec == (uint64_t) -1) + return 0; + + if (usec == 0) + usec = RTNL_DEFAULT_TIMEOUT; + + return now(CLOCK_MONOTONIC) + usec; +} + +static int rtnl_poll(sd_rtnl *rtnl, bool need_more, uint64_t timeout_usec) { + struct pollfd p[1] = {}; + struct timespec ts; + usec_t m = (usec_t) -1; + int r, e; + + assert(rtnl); + + e = sd_rtnl_get_events(rtnl); + if (e < 0) + return e; + + if (need_more) + /* Caller wants more data, and doesn't care about + * what's been read or any other timeouts. */ + return e |= POLLIN; + else { + usec_t until; + /* Caller wants to process if there is something to + * process, but doesn't care otherwise */ + + r = sd_rtnl_get_timeout(rtnl, &until); + if (r < 0) + return r; + if (r > 0) { + usec_t nw; + nw = now(CLOCK_MONOTONIC); + m = until > nw ? until - nw : 0; + } + } + + if (timeout_usec != (uint64_t) -1 && (m == (uint64_t) -1 || timeout_usec < m)) + m = timeout_usec; + + p[0].fd = rtnl->fd; + p[0].events = e; + + r = ppoll(p, 1, m == (uint64_t) -1 ? NULL : timespec_store(&ts, m), NULL); + if (r < 0) + return -errno; + + return r > 0 ? 1 : 0; +} + +int sd_rtnl_wait(sd_rtnl *nl, uint64_t timeout_usec) { + assert_return(nl, -EINVAL); + assert_return(!rtnl_pid_changed(nl), -ECHILD); + + if (nl->rqueue_size > 0) + return 0; + + return rtnl_poll(nl, false, timeout_usec); +} + +static int timeout_compare(const void *a, const void *b) { + const struct reply_callback *x = a, *y = b; + + if (x->timeout != 0 && y->timeout == 0) + return -1; + + if (x->timeout == 0 && y->timeout != 0) + return 1; + + if (x->timeout < y->timeout) + return -1; + + if (x->timeout > y->timeout) + return 1; + + return 0; +} + +int sd_rtnl_call_async(sd_rtnl *nl, + sd_rtnl_message *m, + sd_rtnl_message_handler_t callback, + void *userdata, + uint64_t usec, + uint32_t *serial) { + struct reply_callback *c; + uint32_t s; + int r, k; + + assert_return(nl, -EINVAL); + assert_return(m, -EINVAL); + assert_return(callback, -EINVAL); + assert_return(!rtnl_pid_changed(nl), -ECHILD); + + r = hashmap_ensure_allocated(&nl->reply_callbacks, uint64_hash_func, uint64_compare_func); + if (r < 0) + return r; + + if (usec != (uint64_t) -1) { + r = prioq_ensure_allocated(&nl->reply_callbacks_prioq, timeout_compare); + if (r < 0) + return r; + } + + c = new0(struct reply_callback, 1); + if (!c) + return -ENOMEM; + + c->callback = callback; + c->userdata = userdata; + c->timeout = calc_elapse(usec); + + k = sd_rtnl_send(nl, m, &s); + if (k < 0) { + free(c); + return k; + } + + c->serial = s; + + r = hashmap_put(nl->reply_callbacks, &c->serial, c); + if (r < 0) { + free(c); + return r; + } + + if (c->timeout != 0) { + r = prioq_put(nl->reply_callbacks_prioq, c, &c->prioq_idx); + if (r > 0) { + c->timeout = 0; + sd_rtnl_call_async_cancel(nl, c->serial); + return r; + } + } + + if (serial) + *serial = s; + + return k; +} + +int sd_rtnl_call_async_cancel(sd_rtnl *nl, uint32_t serial) { + struct reply_callback *c; + uint64_t s = serial; + + assert_return(nl, -EINVAL); + assert_return(serial != 0, -EINVAL); + assert_return(!rtnl_pid_changed(nl), -ECHILD); + + c = hashmap_remove(nl->reply_callbacks, &s); + if (!c) + return 0; + + if (c->timeout != 0) + prioq_remove(nl->reply_callbacks_prioq, c, &c->prioq_idx); + + free(c); + return 1; +} + +int sd_rtnl_call(sd_rtnl *nl, + sd_rtnl_message *message, + uint64_t usec, + sd_rtnl_message **ret) { + usec_t timeout; + uint32_t serial; + bool room = false; + int r; + + assert_return(nl, -EINVAL); + assert_return(!rtnl_pid_changed(nl), -ECHILD); + assert_return(message, -EINVAL); + + r = sd_rtnl_send(nl, message, &serial); + if (r < 0) + return r; + + timeout = calc_elapse(usec); + + for (;;) { + usec_t left; + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *incoming = NULL; + + if (!room) { + sd_rtnl_message **q; + + if (nl->rqueue_size >= RTNL_RQUEUE_MAX) + return -ENOBUFS; + + /* Make sure there's room for queueing this + * locally, before we read the message */ + + q = realloc(nl->rqueue, (nl->rqueue_size + 1) * sizeof(sd_rtnl_message*)); + if (!q) + return -ENOMEM; + + nl->rqueue = q; + room = true; + } + + r = socket_read_message(nl, &incoming); + if (r < 0) + return r; + if (incoming) { + uint32_t received_serial = message_get_serial(incoming); + + if (received_serial == serial) { + r = sd_rtnl_message_get_errno(incoming); + if (r < 0) + return r; + + if (ret) { + *ret = incoming; + incoming = NULL; + } + + return 1; + } + + /* Room was allocated on the queue above */ + nl->rqueue[nl->rqueue_size ++] = incoming; + incoming = NULL; + room = false; + + /* Try to read more, right away */ + continue; + } + if (r != 0) + continue; + + if (timeout > 0) { + usec_t n; + + n = now(CLOCK_MONOTONIC); + if (n >= timeout) + return -ETIMEDOUT; + + left = timeout - n; + } else + left = (uint64_t) -1; + + r = rtnl_poll(nl, true, left); + if (r < 0) + return r; + + r = dispatch_wqueue(nl); + if (r < 0) + return r; + } +} + +int sd_rtnl_flush(sd_rtnl *rtnl) { + int r; + + assert_return(rtnl, -EINVAL); + assert_return(!rtnl_pid_changed(rtnl), -ECHILD); + + if (rtnl->wqueue_size <= 0) + return 0; + + for (;;) { + r = dispatch_wqueue(rtnl); + if (r < 0) + return r; + + if (rtnl->wqueue_size <= 0) + return 0; + + r = rtnl_poll(rtnl, false, (uint64_t) -1); + if (r < 0) + return r; + } +} + +int sd_rtnl_get_events(sd_rtnl *rtnl) { + int flags = 0; + + assert_return(rtnl, -EINVAL); + assert_return(!rtnl_pid_changed(rtnl), -ECHILD); + + if (rtnl->rqueue_size <= 0) + flags |= POLLIN; + if (rtnl->wqueue_size > 0) + flags |= POLLOUT; + + return flags; +} + +int sd_rtnl_get_timeout(sd_rtnl *rtnl, uint64_t *timeout_usec) { + struct reply_callback *c; + + assert_return(rtnl, -EINVAL); + assert_return(timeout_usec, -EINVAL); + assert_return(!rtnl_pid_changed(rtnl), -ECHILD); + + if (rtnl->rqueue_size > 0) { + *timeout_usec = 0; + return 1; + } + + c = prioq_peek(rtnl->reply_callbacks_prioq); + if (!c) { + *timeout_usec = (uint64_t) -1; + return 0; + } + + *timeout_usec = c->timeout; + + return 1; +} + +static int io_callback(sd_event_source *s, int fd, uint32_t revents, void *userdata) { + sd_rtnl *rtnl = userdata; + int r; + + assert(rtnl); + + r = sd_rtnl_process(rtnl, NULL); + if (r < 0) + return r; + + return 1; +} + +static int time_callback(sd_event_source *s, uint64_t usec, void *userdata) { + sd_rtnl *rtnl = userdata; + int r; + + assert(rtnl); + + r = sd_rtnl_process(rtnl, NULL); + if (r < 0) + return r; + + return 1; +} + +static int prepare_callback(sd_event_source *s, void *userdata) { + sd_rtnl *rtnl = userdata; + int r, e; + usec_t until; + + assert(s); + assert(rtnl); + + e = sd_rtnl_get_events(rtnl); + if (e < 0) + return e; + + r = sd_event_source_set_io_events(rtnl->io_event_source, e); + if (r < 0) + return r; + + r = sd_rtnl_get_timeout(rtnl, &until); + if (r < 0) + return r; + if (r > 0) { + int j; + + j = sd_event_source_set_time(rtnl->time_event_source, until); + if (j < 0) + return j; + } + + r = sd_event_source_set_enabled(rtnl->time_event_source, r > 0); + if (r < 0) + return r; + + return 1; +} + +static int exit_callback(sd_event_source *event, void *userdata) { + sd_rtnl *rtnl = userdata; + + assert(event); + + sd_rtnl_flush(rtnl); + + return 1; +} + +int sd_rtnl_attach_event(sd_rtnl *rtnl, sd_event *event, int priority) { + int r; + + assert_return(rtnl, -EINVAL); + assert_return(!rtnl->event, -EBUSY); + + assert(!rtnl->io_event_source); + assert(!rtnl->time_event_source); + + if (event) + rtnl->event = sd_event_ref(event); + else { + r = sd_event_default(&rtnl->event); + if (r < 0) + return r; + } + + r = sd_event_add_io(rtnl->event, rtnl->fd, 0, io_callback, rtnl, &rtnl->io_event_source); + if (r < 0) + goto fail; + + r = sd_event_source_set_priority(rtnl->io_event_source, priority); + if (r < 0) + goto fail; + + r = sd_event_source_set_prepare(rtnl->io_event_source, prepare_callback); + if (r < 0) + goto fail; + + r = sd_event_add_monotonic(rtnl->event, 0, 0, time_callback, rtnl, &rtnl->time_event_source); + if (r < 0) + goto fail; + + r = sd_event_source_set_priority(rtnl->time_event_source, priority); + if (r < 0) + goto fail; + + r = sd_event_add_exit(rtnl->event, exit_callback, rtnl, &rtnl->exit_event_source); + if (r < 0) + goto fail; + + return 0; + +fail: + sd_rtnl_detach_event(rtnl); + return r; +} + +int sd_rtnl_detach_event(sd_rtnl *rtnl) { + assert_return(rtnl, -EINVAL); + assert_return(rtnl->event, -ENXIO); + + if (rtnl->io_event_source) + rtnl->io_event_source = sd_event_source_unref(rtnl->io_event_source); + + if (rtnl->time_event_source) + rtnl->time_event_source = sd_event_source_unref(rtnl->time_event_source); + + if (rtnl->exit_event_source) + rtnl->exit_event_source = sd_event_source_unref(rtnl->exit_event_source); + + if (rtnl->event) + rtnl->event = sd_event_unref(rtnl->event); + + return 0; +} + +int sd_rtnl_add_match(sd_rtnl *rtnl, + uint16_t type, + sd_rtnl_message_handler_t callback, + void *userdata) { + struct match_callback *c; + + assert_return(rtnl, -EINVAL); + assert_return(callback, -EINVAL); + assert_return(!rtnl_pid_changed(rtnl), -ECHILD); + assert_return(message_type_is_link(type) || message_type_is_addr(type) || message_type_is_route(type), -ENOTSUP); + + c = new0(struct match_callback, 1); + if (!c) + return -ENOMEM; + + c->callback = callback; + c->type = type; + c->userdata = userdata; + + LIST_PREPEND(match_callbacks, rtnl->match_callbacks, c); + + return 0; +} + +int sd_rtnl_remove_match(sd_rtnl *rtnl, + uint16_t type, + sd_rtnl_message_handler_t callback, + void *userdata) { + struct match_callback *c; + + assert_return(rtnl, -EINVAL); + assert_return(callback, -EINVAL); + assert_return(!rtnl_pid_changed(rtnl), -ECHILD); + + LIST_FOREACH(match_callbacks, c, rtnl->match_callbacks) + if (c->callback == callback && c->type == type && c->userdata == userdata) { + LIST_REMOVE(match_callbacks, rtnl->match_callbacks, c); + free(c); + + return 1; + } + + return 0; +} diff --git a/src/libsystemd/test-rtnl.c b/src/libsystemd/test-rtnl.c new file mode 100644 index 0000000000..58654e990a --- /dev/null +++ b/src/libsystemd/test-rtnl.c @@ -0,0 +1,330 @@ +/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ + +/*** + This file is part of systemd. + + Copyright 2013 Tom Gundersen <teg@jklm.no> + + systemd is free software; you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published by + the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + systemd is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with systemd; If not, see <http://www.gnu.org/licenses/>. +***/ + +#include <linux/rtnetlink.h> +#include <netinet/ether.h> + +#include "util.h" +#include "macro.h" +#include "sd-rtnl.h" +#include "socket-util.h" +#include "rtnl-util.h" +#include "event-util.h" + +static void test_link_configure(sd_rtnl *rtnl, int ifindex) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *message; + uint16_t type; + const char *mac = "98:fe:94:3f:c6:18", *name = "test"; + unsigned int mtu = 1450; + void *data; + + /* we'd really like to test NEWLINK, but let's not mess with the running kernel */ + assert(sd_rtnl_message_link_new(RTM_GETLINK, ifindex, &message) >= 0); + assert(sd_rtnl_message_append_string(message, IFLA_IFNAME, name) >= 0); + assert(sd_rtnl_message_append_ether_addr(message, IFLA_ADDRESS, ether_aton(mac)) >= 0); + assert(sd_rtnl_message_append_u32(message, IFLA_MTU, mtu) >= 0); + + assert(sd_rtnl_message_read(message, &type, &data) > 0); + assert(type == IFLA_IFNAME); + assert(streq(name, (char *) data)); + + assert(sd_rtnl_message_read(message, &type, &data) > 0); + assert(type == IFLA_ADDRESS); + assert(streq(mac, ether_ntoa(data))); + + assert(sd_rtnl_message_read(message, &type, &data) > 0); + assert(type == IFLA_MTU); + assert(mtu == *(unsigned int *) data); + + assert(sd_rtnl_call(rtnl, message, 0, NULL) == 1); +} + +static void test_route(void) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *req; + struct in_addr addr; + uint32_t index = 2; + uint16_t type; + void *data; + int r; + + r = sd_rtnl_message_route_new(RTM_NEWROUTE, AF_INET, &req); + if (r < 0) { + log_error("Could not create RTM_NEWROUTE message: %s", strerror(-r)); + return; + } + + addr.s_addr = htonl(INADDR_LOOPBACK); + + r = sd_rtnl_message_append_in_addr(req, RTA_GATEWAY, &addr); + if (r < 0) { + log_error("Could not append RTA_GATEWAY attribute: %s", strerror(-r)); + return; + } + + r = sd_rtnl_message_append_u32(req, RTA_OIF, index); + if (r < 0) { + log_error("Could not append RTA_OIF attribute: %s", strerror(-r)); + return; + } + + assert(sd_rtnl_message_read(req, &type, &data) > 0); + assert(type == RTA_GATEWAY); + assert(((struct in_addr *)data)->s_addr == addr.s_addr); + + assert(sd_rtnl_message_read(req, &type, &data) > 0); + assert(type == RTA_OIF); + assert(*(uint32_t *) data == index); +} + +static void test_multiple(void) { + sd_rtnl *rtnl1, *rtnl2; + + assert(sd_rtnl_open(0, &rtnl1) >= 0); + assert(sd_rtnl_open(0, &rtnl2) >= 0); + + rtnl1 = sd_rtnl_unref(rtnl1); + rtnl2 = sd_rtnl_unref(rtnl2); +} + +static int link_handler(sd_rtnl *rtnl, sd_rtnl_message *m, void *userdata) { + void *data; + uint16_t type; + char *ifname = userdata; + + assert(rtnl); + assert(m); + + log_info("got link info about %s", ifname); + free(ifname); + + while (sd_rtnl_message_read(m, &type, &data) > 0) { + switch (type) { +// case IFLA_MTU: +// assert(*(unsigned int *) data == 65536); +// break; +// case IFLA_QDISC: +// assert(streq((char *) data, "noqueue")); +// break; + case IFLA_IFNAME: + assert(streq((char *) data, "lo")); + break; + } + } + + return 1; +} + +static void test_event_loop(int ifindex) { + _cleanup_event_unref_ sd_event *event = NULL; + _cleanup_sd_rtnl_unref_ sd_rtnl *rtnl = NULL; + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *m = NULL; + char *ifname; + + ifname = strdup("lo2"); + assert(ifname); + + assert(sd_rtnl_open(0, &rtnl) >= 0); + assert(sd_rtnl_message_link_new(RTM_GETLINK, ifindex, &m) >= 0); + + assert(sd_rtnl_call_async(rtnl, m, &link_handler, ifname, 0, NULL) >= 0); + + assert(sd_event_default(&event) >= 0); + + assert(sd_rtnl_attach_event(rtnl, event, 0) >= 0); + + assert(sd_event_run(event, 0) >= 0); + + assert(sd_rtnl_detach_event(rtnl) >= 0); +} + +static int pipe_handler(sd_rtnl *rtnl, sd_rtnl_message *m, void *userdata) { + int *counter = userdata; + + (*counter) --; + + log_info("got reply, %d left in pipe", *counter); + + return sd_rtnl_message_get_errno(m); +} + +static void test_async(int ifindex) { + _cleanup_sd_rtnl_unref_ sd_rtnl *rtnl = NULL; + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *m = NULL, *r = NULL; + uint32_t serial; + char *ifname; + + ifname = strdup("lo"); + assert(ifname); + + assert(sd_rtnl_open(0, &rtnl) >= 0); + + assert(sd_rtnl_message_link_new(RTM_GETLINK, ifindex, &m) >= 0); + + assert(sd_rtnl_call_async(rtnl, m, &link_handler, ifname, 0, &serial) >= 0); + + assert(sd_rtnl_wait(rtnl, 0) >= 0); + assert(sd_rtnl_process(rtnl, &r) >= 0); +} + +static void test_pipe(int ifindex) { + _cleanup_sd_rtnl_unref_ sd_rtnl *rtnl = NULL; + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *m1 = NULL, *m2 = NULL; + int counter = 0; + + assert(sd_rtnl_open(0, &rtnl) >= 0); + + assert(sd_rtnl_message_link_new(RTM_GETLINK, ifindex, &m1) >= 0); + assert(sd_rtnl_message_link_new(RTM_GETLINK, ifindex, &m2) >= 0); + + counter ++; + assert(sd_rtnl_call_async(rtnl, m1, &pipe_handler, &counter, 0, NULL) >= 0); + + counter ++; + assert(sd_rtnl_call_async(rtnl, m2, &pipe_handler, &counter, 0, NULL) >= 0); + + while (counter > 0) { + assert(sd_rtnl_wait(rtnl, 0) >= 0); + assert(sd_rtnl_process(rtnl, NULL) >= 0); + } +} + +static void test_container(void) { + _cleanup_sd_rtnl_message_unref_ sd_rtnl_message *m = NULL; + uint16_t type; + void *data; + + assert(sd_rtnl_message_link_new(RTM_NEWLINK, 0, &m) >= 0); + + assert(sd_rtnl_message_open_container(m, IFLA_LINKINFO) >= 0); + assert(sd_rtnl_message_open_container(m, IFLA_LINKINFO) == -EINVAL); + assert(sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "kind") >= 0); + assert(sd_rtnl_message_close_container(m) >= 0); + assert(sd_rtnl_message_close_container(m) == -EINVAL); + + assert(sd_rtnl_message_read(m, &type, &data) == -EINVAL); + +/* TODO: add support for entering containers + assert(sd_rtnl_message_read(m, &type, &data) > 0); + assert(type == IFLA_INFO_KIND); + assert(streq("kind", (char *) data)); + + assert(sd_rtnl_message_read(m, &type, &data) == 0); +*/ +} + +static void test_match(void) { + _cleanup_sd_rtnl_unref_ sd_rtnl *rtnl = NULL; + + assert(sd_rtnl_open(0, &rtnl) >= 0); + + assert(sd_rtnl_add_match(rtnl, RTM_NEWLINK, &link_handler, NULL) >= 0); + assert(sd_rtnl_add_match(rtnl, RTM_NEWLINK, &link_handler, NULL) >= 0); + + assert(sd_rtnl_remove_match(rtnl, RTM_NEWLINK, &link_handler, NULL) == 1); + assert(sd_rtnl_remove_match(rtnl, RTM_NEWLINK, &link_handler, NULL) == 1); + assert(sd_rtnl_remove_match(rtnl, RTM_NEWLINK, &link_handler, NULL) == 0); +} + +int main(void) { + sd_rtnl *rtnl; + sd_rtnl_message *m; + sd_rtnl_message *r; + void *data; + int if_loopback; + uint16_t type; + unsigned int mtu = 0; + unsigned int *mtu_reply; + + test_match(); + + test_multiple(); + + test_route(); + + test_container(); + + assert(sd_rtnl_open(0, &rtnl) >= 0); + assert(rtnl); + + if_loopback = (int) if_nametoindex("lo"); + assert(if_loopback > 0); + + test_async(if_loopback); + + test_pipe(if_loopback); + + test_event_loop(if_loopback); + + test_link_configure(rtnl, if_loopback); + + assert(sd_rtnl_message_link_new(RTM_GETLINK, if_loopback, &m) >= 0); + assert(m); + + assert(sd_rtnl_message_get_type(m, &type) >= 0); + assert(type == RTM_GETLINK); + + assert(sd_rtnl_message_read(m, &type, &data) == 0); + + assert(sd_rtnl_call(rtnl, m, 0, &r) == 1); + assert(sd_rtnl_message_get_type(r, &type) >= 0); + assert(type == RTM_NEWLINK); + + assert(sd_rtnl_message_read(m, &type, &data) == 0); + assert((r = sd_rtnl_message_unref(r)) == NULL); + + assert(sd_rtnl_call(rtnl, m, -1, &r) == -EPERM); + assert((m = sd_rtnl_message_unref(m)) == NULL); + assert((r = sd_rtnl_message_unref(r)) == NULL); + + assert(sd_rtnl_message_link_new(RTM_GETLINK, if_loopback, &m) >= 0); + assert(m); + + assert(sd_rtnl_message_append_u32(m, IFLA_MTU, mtu) >= 0); + assert(sd_rtnl_message_read(m, &type, (void **) &mtu_reply) == 1); + + assert(type == IFLA_MTU); + assert(*mtu_reply == 0); + + assert(sd_rtnl_message_read(m, &type, &data) == 0); + + assert(sd_rtnl_call(rtnl, m, -1, &r) == 1); + while (sd_rtnl_message_read(r, &type, &data) > 0) { + switch (type) { +// case IFLA_MTU: +// assert(*(unsigned int *) data == 65536); +// break; +// case IFLA_QDISC: +// assert(streq((char *) data, "noqueue")); +// break; + case IFLA_IFNAME: + assert(streq((char *) data, "lo")); + break; + } + } + + assert(sd_rtnl_flush(rtnl) >= 0); + + assert((m = sd_rtnl_message_unref(m)) == NULL); + assert((r = sd_rtnl_message_unref(r)) == NULL); + assert((rtnl = sd_rtnl_unref(rtnl)) == NULL); + + return EXIT_SUCCESS; +} |